X-ray image similarity estimation using triplet neural networks

In this research triplet neural network is evaluated for use as an image quality assessment method. Quality is assessed by comparing image with more structural deformations with image with fewer structural deformations. The more similar they are, the better the quality of deformed image. The neural network model is compared with mean square error, structural similarity and multi-scale structural similarity quality assessment methods. The neural network was trained to assess the quality of low dose CT images. The experiment showed that triplet neural network is suitable for image quality assessment and that it strongly correlates with structural similarity and multi-scale structural similarity quality assessment methods